Slashdot Mirror
400 Gigabits Per Square Inch
NWprobe writes "Some scientists at Naval Research Laboratory have developed a new super disk. Nando Times has an article about it. I want a storage device like this, but will we ever see them come into production...
" "We anticipate we can put 400 gigabits in a square inch," said solid state physicist Gary Prinz of the Naval Research Laboratory, which has just contracted with a pioneering Minnesota firm to move the technology from the lab to the production line. "
Back in the days of 80 ns memory chips, the best bubble memory chips were something like 400 ns (.4 ms) as I recall. This would make them about 10x faster than the best discs today. The problem was with the magnetic domain and RLC resonances frequencies--with the "on insulator" type of tricks that are going on today, a real speed-up might be possible. Unfortunately, all this seems to have been tied up in small companies with only one customer (the US military), and there probably isn't the money to push their performance with heavy duty development efforts. There may have been some minimum size problems with the technology, so you couldn't get the densities you really would like without a lot of bucks on that too.
This is not a question of increasing the storage capacity of regular size devices (mostly). It is a case of being able to store the same amount of information in an incredibly smaller area using much less power.
Applications:
digital cameras
wearables
smart appliances
CPU cache! (mentioned in article)
laptops
I'm sure people could think of more...
-- perl -e'print pack"H*","6e656d6f406d38792e6f7267"'
Smaller, cooler, more efficient. I'll take some.
The mystery company is almost doubtlessly Seagate, who's headquarters is based right next door to where I work.
But, what I find interesting about this particular miracle is the possibility of putting a few gigs of storage on the same chip as the CPU. Probably not very practical for general-purpose computers (wanna reload your data just so you can have a faster CPU?), but there are other uses...
Hook a wee bit of this to the equivalent of a 386 or 486, put it in an affordable package, and you could have:
- hyper-intelligent watch/notetaker
- key-fob sized MP3 player
- standard-sized 5x7 picture frame that changes pictures every hour to show a different image of the wife and kids
- electronic books
- those flying balls inspired by the thing from Star Wars
- smart appliances (oh no, not again!)
and so on...Of course, it all depends on the final cost.
Stupid people will be persecuted to the fullest extent allowed by law.
This is solid-state storage, not a new type of hard disk, so it's square inch of silicon rather than square inch of disk surface.
This appears to be the latest in a long line of "hard disk killers". Remember bubble memory? E2PROM? FLASH? Who knows, maybe this one will be more successful than those. I'm not holding my breath...
Thankfully, defragging happens in O(n) time, meaning that if you triple the hard drive space, you triple the time it takes to defrag. It's not like sorting O(n log(n)), which would get monstorous.
Factoring in access and write speed, it's actually O(n/s) where s is the read/write/access speed.
Basically, if you have a 2Tb drive that's 100 times as fast as your 20Gb drive, it'll take exactly the same amount of time to defrag, though if you wanted to sort it at the same time, it would take 7 times longer (ln 100 ~= 7.5).
Kevin Fox
Kevin Fox
We all keep reading articles on new, cheap, large storage, but when will we all actually get to buy them?
Was it me, or did he imply that some of this stuff could be dumped on an area of a CPU, essentially giving you an on-die disk? Seek times measured in clock cycles? Yum.
--
E_NOSIG
Any specs on those? I mean, the technology sounds great, but how *fast* is MRAM/VRAM?
engineers never lie; we just approximate the truth.
I doubt that speed is going to be an issue.
;-)
Cool.
If speed isn't an issue, then price is. If this is faster or as fast as SDRAM, it will replace SDRAM.
However, I don't think these will ever compete with traditional magnetic harddrives. These will be VERY expensive in comparison, and only people who are willing to spend lots of money for (physical) stability (it's Solid State, ie. no moving parts) and possibly faster access time. So maybe big servers could use this, but, only if it's less expensive than RAID (but RAID has redundancy...) or if the server is located inside a paint mixing machine...
SO, sure, this will replace/compete with NVRAM.
And depending (a lot!) on price, it may replace SDRAM.
But I doubt if it will replace physical (platter) hard drives (which, I should add, they are not claiming in the article...).
I will say, though, that if they can get the price down to anywhere near traditional drives (now, about $10 per GB or less) then this will replace hard drives!
What about fsck()?
I can throw myself at the ground, and miss.
For just the frame with the story in it, click here.
--
Time is Nature's way of keeping everything from happening at once... the bitch.
--
Time is Nature's way of keeping everything from happening at once... the bitch.
Well, imagine if this stuff was as fast as (or even faster than) memory is now. This stuff could be the "memory" and the "hard drive" and the "CPU cache"... or maybe there wouldn't be a difference anymore. Having a CPU attached to a square inch of that stuff means you have a CPU with memory and nonvolitle storage that equals 40GB... that would pave the way for more lower cost PCs.
Besides... saying that this technology wouldn't be useful and no one would care is just like saying that no one will ever need more that 640k RAM.
--
"What do you want me to do? Whack a guy? Off a guy? Whack off a guy? Cause I'm married."
Where am I gonna get that kind of power Tom?!
---
DO NOT DISTURB THE SE
I am curious what the state change rate is (how fast it can be changed from a 1 to a 0) and if I can change any of the bit storage locations on the device simoultaneously or one at a time...
The rate is important, for instance memory can chage 100*10^9 bits per second (theoretically) for PC100 RAM (100 MHz). Hard disks can change 66*10^9 bits per second (ATA66), but only the bit the hard disk head is over at that time.
This is very oversimplified (all throughputs are absolute theoretical maximums), and I probably mesed up the exponents, and someone more knowledgable can refine my question...but hopefully I got enough of a point across for someone to understand my question--is this tech really feasible as a memory replacement, and where is there more hard data about it?
- Sig
Actually, If the platters are about 3.5" in diameter (1.75" in radius), and there's about 1" in diameter lost in the middle, or .5" in radius, the area on this disk would be *(1.75^2) - *(.5^2) = 8.835 in^2 * 50 = 441.75 * (2*2) = 1767 or 1.767 TB
"I like systems, their application excepted", George Sand (French)
Interestingly enough, I recall working on electronics repair training equipment in the Navy in the late 80's that used this kind of memory. This memory was part of a computer (based on the Z80 I think, it's been a long time ) If you looked at it under a powerful magnifier of some sort you could actually see little round ferrite cores wrapped in extraordinarily fine copper wire. I always wondered how anybody managed to wrap that in a production environment.
"We apologize for the inconvenience."
yes Seagate has a big campus here, like 6 buildings or so, but its actually headquartered in California. 3M on the other hand is headquartered over in St. Paul, and their Imation division has broken off from the parent company, and has been doing vast amounts of research on storage devices. my guess would be 3M and Imation over Seagate. but i could totally be wrong.
That is what Bill Gates was quoted back in the 80's. I am running 256MB ram right now. My point? I am sure that 50000GB seems quite insane right now, it may be the minimum reqs for Windows 2005.
In fact, you make very little sense. For a PalmTop application, this technology would yield a sufficient storage space in a very small footprint. Gee, isn't that what we want?
I do not see how you think that this would be disaster in the marketplace. (What would you have thought of a 10GB drive back in the 80's?)
I couldn't say I wasn't impressed about this. Especially because I know the hype 20 years ago when the first hdd appeared. It had about 5Mb and everyone was wondering who could ever fill this huge space.
Yet, let's not be too enthusiastic because for the moment this technology has some possible problems (at least in what concerns the military - which seem to be the first interested in it). Being a magnetic storage device a magnetic field can destabilize the data on it, thus making the computer unusable (or unreliable). Hypothetically speaking, it would be quite easy to attack a ship with some sort of radiation that will make it vulnerable to another attack that may destroy it
On the other hand, civilian users are quite protected against this, since there aren't many important secrets to be destroyed (I'm not speaking about corporate users). Then again, who will ever use 400Gb (or more) at home ? (ok I may sound silly and repeating the sentence I mentioned in the beginning). There are some limits in one's ability to gather information and one of them is the time.
Although, if I were to speak frankly, I imagine Windows 2010 taking 70% of this space (with a 25Gb Solitaire). Hopefully there will be no windows any more in 2010
Well if I remember back to my digital classes, there are going to have to be ALOT of address lines to access each individual bit, which means more metal, which means more heat! Especially if they say they are going to try and stack them (like poker chips). Also, I can see why the manufacturing process would be cheaper. As the article says, there are NO transistors. If anyone has taken a VLSI class they know how many steps are involved in the fabrication process of transistors....Its a wonder how intel and amd can sell their chips for so little!
People have commented more room for there .mp3's.
I will hope I longer to need a compresion format for my music. I prefer Quality.
The Kruger Dunning explains most post on
Let me guess, you're a home user. No you won't have use for this large a storage device in the near future but many network admins work with much larger storage devices already. And while I can't speak for others, this certainly has me interested. I've had to manage as much as 240 GBs of data when 10 GB hard drives were the biggest we could get out hands on. It would have made my life a lot easier to have all that data on one drive rather than spread out on 24 (more when you count the 'lost' drives in the RAIDs).
Additionally, you have to remember this is a SOLID STATE drive. No moving parts, the tranfer times will be almost as fast as your RAM or cache. While this may not seem like that big a deal to a home user, this type of speed will allow developers (Web and other) to do more processing on their own systems and then transfer just the results to your computer. Which means better content can be delievered to the customer. Do I see a immersive 3D Slashdot environment in our future?......maybe.......lol
Dave
My memory is that bubble memory involved the creation of bubble of charge within the substrate by the application of a magnetic field. These bubbles could be polarised in some way to indicate 0 or 1. The size of these bubbles depended upon the strength of the magnetic field with very small bubbles requiring very strong fields (just what you want in a computer).
Also I think that it was a linear storage medium where the bubbles were moved in a loop around the chip by placing voltages on T shaped elements. This would probably mean that the speed of retrieval is not very great, as the bubbles are led past a read-out area.
It probably was just a technology that did not offer any advantage over easier or already existing devices. One advantage was meant to be non-volatility (with a battery backup). Now we have Flash memory.
Gamma Testing - Where testing is extended to the full user community (AKA Shipping the Program)
*thinking*If i start defrag today...will it be finished till i return from vacation?
a signature
Anyway, back in reality, I'll stick to getting excited about actual product shipments.
What did you eat today? http://www.atetoday.com/
For Nonvolatile Electronics, Inc.'s site, go to http://www.nve.com. They have some white papers on Magnetoresistive Random Access Memory (MRAM). Unlike that "alien technology" we've seen posted before, this looks to promise truly massive nonvolatile storage.
let's see 400/8 = 50 gigabytes per square inch...given that platters for a 3.5 inch drive are...~4 inches....pir^2 = 50 square inches in area....50 * 50 = 2500 gigabytes...times 2 sides... 5000 gigabytes per platter...times 2 platters a hard drive....10000 gigabytes on your standard harddrive....no?
:)
someone please double check that
It seems as though this could be built right into the cpu or into the package holding the cpu die, which would lead to getting your hard drive, ram, and cpu all in one unit.
So what happens when that package, with a certain company's operating system permanently installed and hardwired in, is available cheaper than the same hardware without any os, or with a certain "free" os pre-installed?
If this thing is "instant-on", then either os will be right there, ready to go as soon as you hit the switch, but one certain company will be able to subsidise the purchase, whereas the other os, even though free, won't have a financial behemoth behind it (unless it's AOL, and you have to be a subscriber of theirs to get the discount, or maybe even for the hardware to work at all).
I see even classic Slashdot is now pretty much unusable on dial up anymore.
Nando Times is a spin-off of the Raleigh, North Carolina newspaper "The News and Observer". The N&O (n and o, nando, get it?) got into the ISP business several years ago with, IIRC, Nando.net, which I think still exists, but the actual subscriber base got sold off to Mindspring a few years back.
If linking like that becomes illegal, won't that pretty much be the end of the internet as we know it?
As to whether or not Slashdot can or should provide content, I don't think that's what it's really here for. We the Slashdot audience provide the content. Unfortunately the content is sometimes of the quality of posts like yours.
I see even classic Slashdot is now pretty much unusable on dial up anymore.
Heh.. I recall reading on New Scientist about using really small pins dragged around a wax disk or cylinder or something. You could heat it up to make a pit, and if it hit a pit it would heat up slightly due to friction while getting out of it and you could detect that. Those pits could be really damn small.
I tried a search on their website, but couldn't come up with a link, could be its only in their print version (and I'm not searching THAT!) or I just wasn't lucky.
-W
IIRC those chips were made out of a thin layer of garnet, and the description of the individual memory cells being shaped like "tiny doughnuts" rings a bell. At the time I think the biggest one they had was about 500K, but considering how much smaller the current electronic paths are in state of the art semi-conductors compared to what was available in 1986), I find myself wondering how this "new" technology is different from the older one. Is anyone out there in /. land familiar with both enough to fill in the details?
...Open Source isn't the only answer -- but it's almost always a better value than the alternatives...
I think the key to this technology really is the size. The problem with PC's now is that even though the processor is getting faster, the rest of the system is lagging behind. With faster and smaller memory, and the ability to put the memory right on the CPU, we could really see the speed boost we are looking for.
A major advantage of the new technology is that the memory system is nonvolatile
Kiss your slow hard drives goodbye, and now almost all of the computer can fit on one card. All the case is needed for now it the expansion slots. Too bad it is too early in development, though...
The thing that matters is the balance between size, speed, and price. The article doesn't seem to touch upon this point very much.
There are many very promising technologies out there. The real problem with memory right now does not seem to be the size at all. For example, hard disk density (size) doubles every year. However, the access time (speed) only decreases about 30% every 10 years.
I think these figures are correct. If not please correct me.
First of all, the technology promises 400 gigabits/in^2, which translates to 50 GB. Don't even bother to translate how much it would be if you made a hard drive out of it, because this technology will push us far beyond hard drives. Imagine something like Sony's memory sticks with 50 GB on it. You could throw about 10 DVDs on it. Buisiness cards, with 100 GB of storage on it. Little MP3 players with years of listening time. Hell, we could probably just have portable WAV players. This will be particularly interesting when compact computer parts like monitors become viable. If I could have a pair of glasses with a display, and a tiny DVD player the size of a pager, with piles of DVDs stored on tiny memory sticks, this would be ideal, and not too far off. You then unplug you display glasses from the DVD player and plug it in your wearable computer, and have access to a huge amount of pawer and data in a very small size. This brings us much closer to wearable and micro-portable electronics.
"I like systems, their application excepted", George Sand (French)
10,000 Gigabits in a volume size of a sugar cube storage device made at the University of Toronto by Dr. Eugenia Kumacheva working at UofT and sponsored by XEROX.
http://www.geocities.com/roman_mi r.geo/NoFCFS.gif
and this is what I have to say about that.
You can't handle the truth.
It really did have to happen, but it still kicks ass. They talk in the article about how the entire system is digital - i.e. you don't have a spinning disk or any mechanical pieces. You take out the mechanical pieces, and you've eliminated most of the reason drives fail. Not only that, but the whole reason hard drives are so slow as compared to say, processor registers or RAM is that they're rotational - they're mechanical and not digital.
:) They say that the two characteristics of storage, speed and price, are roughly proportional - the less you pay the slower it is.
Of course, ideally, you could have a mass storage unit that was several gigabytes of the same stuff that processor registers are made of, but I wouldn't even care to know how much one of those drives would cost.
I'm just wondering how long it will take to get one of these to market. I wish Moore's law also applied to time-to-market.
-- Truth goes out the door when rumor comes innuendo. -- Groucho Marx
Because, dimwit, from the engineering side, it's ALWAYS bits.
Memory chips are measured in BITS. We put a bunch of them together (a couple 4x128mbit's ) to make one megaBYTE, for you, the consumer who wouldn't understand bits.
Bits are more accurate. More quantifiable.
Traditionally, Kilobyte refers to 1000 bytes when dealing with data transmission, and 1024 bytes when dealing with memory.
Now, when dealing with software written by those who don't know this, it could be either.
But a kilobit is always a kilobit. 1000 bits.
Ethernet is 100 MegaBIT because it's channel usage is measured in bits Things go on and off it a bit at a time. and NOT always in even increments of 8. Same for gigabit.
The SAME FOR THE WAY hard drives encode data! What is actually stored on the drive has little directly to do with what you think is stored there. All kinds of encoding is used. Each bit may be comprised of three bits on the platter....
Talk about computing coming full cicle. This sounds like nano-scale magnetic core memory.
What's next? 0.6 micron punch cards?
--
then it comes to be that the soothing light at the end of your tunnel is just a freight train coming your way
If the per-chip cost is low enough, it would probably be more cost effective to surface-mount the components and chuck the whole unit when one went bad than it would be to have a socketed board and replace an individual chip. For a consumer product, this approach would probably work well -- when one chip died, the whole unit would keep working, giving the user time to buy a new unit and back up his data onto the new one. Considering that the MTBF on solid-state electronics is pretty high, by the time one chip died the whole unit would be effectively obsolete anyway. For a server, you'd want it to be hot-swappable; but for a consumer product a sealed unit would be fine.
"The axiom 'An honest man has nothing to fear from the police'
Why is it that the proponents of "one nation under God" are so eager to get rid of "liberty and justice for all"?
Microsoft announced that it was now able to ship beta copies of Windows 2002 on a single, 8" hard drive unit. Under further questioning, the spokesperson admitted that they had attempted to use the latest AI compression algorithms to remove redundancy, but that the program had simply wiped the disks clean. The spokesperson later said that, as yet, it had not been determined if this had been a bug in the AI.
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
I really thought the whining would stop, but instead both users and the industry have chosen to ignore the new prefixes (summarized below to emphasize the triviality of quibbling) I did not expect everyone to start doing instant conversions, but I did expect them to start using the units as a ballpark indicator of which sort of 'mega' they were using.
True, the difference between terabit and tebibit is only 10%, but if you're going to whine about that 10% (or the 5% megabit gap), presumably you should be using the new standards.
__________
If you can go to bed, knowing you did a valuable thing today, you're very lucky. If you can't... it's not bedtime
More importantly, because there are no moving parts, you'd have incredible levels of reliability and very low latency. This is sorely needed -- while hard drive capacity has been advancing rapidly, hard drive speed has only made modest improvements. In many applications (databases, for example) the biggest performance bottleneck is physical I/O. Even with the fastest hard drives, you still have latency measured in milliseconds (10^-3) -- because you have to wait while you wait for the platter to spin around to the byte you need (on a 10k RPM drive, you have to wait an average of 12ms to [physically] read an arbitrary sector). Conventional RAM, with nanosecond (10^-9) level latency, is 6 orders of magnitude faster -- that's roughly 1 million times faster, for the math-challenged. Getting rid of this disparity has enormous significance for I/O intensive computing.
The real implication here isn't having a multi-terabyte hard drive on your desktop, but having hard drives that actually keep up with the rest of the computer.
"The axiom 'An honest man has nothing to fear from the police'
Why is it that the proponents of "one nation under God" are so eager to get rid of "liberty and justice for all"?